Short Paper Showcase: Group C
Date: Monday 2 Nov 2026
Time: 16:40–17:00
Add this session to your calendar. Talks run back-to-back in the order below.
| Order | Presentation | Abstract | Authors | Contribution type |
|---|---|---|---|---|
| 1 | Mapping the Landscape of Non-Verbal Communication Tracking in Videoconferencing: A Systematic Literature Review | Non-verbal cues help people communicate emotion, attention, and intent, but are harder to perceive in videoconferencing because of restricted framing, camera use, and audio-visual limitations. We present a systematic review of 35 papers on systems that track, interpret, or represent non-verbal communication cues in videoconferencing. We map how these systems capture cues, present outputs, and are evaluated. Most rely on camera and microphone inputs, provide real-time feedback, and use bespoke labels or representations. Evaluations commonly use controlled studies and convenience samples, with limited participatory design and little attention to users with disabilities. We derive a minimum reporting structure for cue-to-feedback design and identify priorities for accountable labelling, ecological and participatory evaluation, and accessible, user-controlled systems. | Vazquez Gonzalez, Neate, Borgo | Short paper |
| 2 | Generative AI as Communication Prosthesis: Design Tensions and Principles for Adaptive Support in Apraxia of Speech | Apraxia of Speech (AoS) is a motor speech disorder characterised by inconsistent articulation, disrupted prosody and effortful initiation. Existing augmentative and alternative communication (AAC) tools and emerging AI-driven speech systems treat the speaker as a stable signal source, yet AoS produces unstable, variable input that breaks this assumption. This short paper reframes generative AI not as a productivity tool nor as an AAC accelerator, but as a communication prosthesis: an adaptive partner that infers intent from unreliable speech while preserving the user's agency, voice and trust. We articulate four HCI design tensions specific to AoS (input variability vs. response consistency; agency preservation vs. assistance density; fluent assistance vs. calibrated uncertainty; adaptive autonomy vs. explainable control) and propose five design principles for adaptive AI support. The paper aims to provoke conversation in the HCI community about how large language models should behave when the user's voice is itself the contested object. | Omoyemi | Short paper |
| 3 | Adaptive Conversational Agents for Motivation and Engagement in Task-Related Goal Planning Dialogues | The psychological quality of a conversational agent’s dialogue style shapes user engagement and experience, yet controlled experimental evidence for motivational effects remains limited, particularly for brief task-related single-session interactions. This study presents a within-subjects experiment examining whether two Self-Determination Theory-grounded dialogue styles, a warm, autonomy-supportive Friend condition and a structured, competence-oriented Coach condition, can enhance task-related state motivation (users’ in-session drive to pursue their stated goal) and user engagement compared to a non-adaptive Generic baseline condition. All three conditions employed identical session-scoped conversational memory, isolating dialogue style as the sole manipulated variable. The evaluation employed a mixed-method approach combining validated self-report scales, exploratory behavioural indicators and qualitative thematic analysis. Both adaptive conditions statistically significantly outperformed the SDT-neutral Generic baseline in this controlled comparison in terms of improved engagement and motivation (both with large measured effect sizes). The two SDT styles did not differ significantly, though preference, behavioural and qualitative indicators converge tentatively on competence support mattering more than relatedness in task-focused planning. These findings provide empirical evidence that psychologically informed dialogue design enhances engagement and motivation in brief goal-oriented interactions, with practical implications for conversational AI design. | Yu, Robb | Short paper |
| 4 | Training Alone Together: Developing a Virtual Opponent for Home-Based Kickboxing Training | Solo training leaves martial artists without a resource it cannot replace: a responsive sparring partner - an absence the COVID-19 pandemic made acute when it closed gyms and dojos worldwide. Replicating that partner digitally requires real-time recognition of a practitioner’s movement, which is still an open challenge for SportsHCI. Virtual Reality (VR) is one route: consumer headsets can place a digital opponent in front of the practitioner, though few actually adapt to how the user is moving. To address this gap, drawing from our experiences as martial arts practitioners, we present a proof-of-concept VR sparring system for kickboxing that attempts to close a feedback loop of recognition, prediction and response for a solo practitioner. Webcam pose estimation feeds a machine learning (ML) classifier, whose output drives an ML sequence predictor that selects the response of an avatar opponent in Unity, a 3D game engine. The classifier reached 87.22% test accuracy across 18 movement classes but reliably recognised only three in deployment. Assessed through autobiographical design, the system is presented as an early exploration from which we draw two provisional design principles for responsive digital sparring partners: calibrated unpredictability and a readable opponent. | McCrossan, Stawarz | Short paper |
| 5 | Trust Me, I’m Human! Comparing Attitudes Towards AI versus Human Junior Colleagues in Medical Device Design | Regulatory demands for transparency and accountability conflict with opaque artificial intelligence (AI) in the medical device design industry. A mixed-methods design was used to study how trust shapes human-AI roles, where interviews with design professionals (N=7) highlighted the importance of safety-criticality in this context. This informed a 2x2 within-subjects experiment (N=24) that manipulated information source (AI, junior colleague) and task safety-criticality (high, low). Crucially, participants were presented with similar reports but were told they were created by either a junior colleague or an AI. Significantly higher trust ratings were given for reports purportedly written by colleagues than for AI. An interaction was found for perceived success, with AI rated significantly higher in non-safety-critical tasks, but not in safety-critical tasks. The results suggest that in regulated industries, where trust depends on accountability rather than just ability, efficiency is prioritised until safety is jeopardised. | Chuah, Lawson | Short paper |
| 6 | Designing for Informal Mindfulness: Exploring How Interactive Technology Can Support Mindful Awareness in Everyday Life | Informal mindfulness, the practice of bringing mindful awareness to everyday activities, has been associated with a range of well-being benefits, yet most technologies focus on formal practices (e.g., meditation). We explore how interactive technology can support informal mindfulness. We conducted interviews with informal mindfulness practitioners (n=12) and used the findings to design a tangible prototype, the Mindfulness Ball. We then evaluated it in a day-long in-the-wild study with 12 participants, followed by exit interviews. Our findings show that the Mindfulness Ball can support informal mindfulness through low-threshold, embodied interactions that prompt brief moments of awareness during daily activities. Participants reported increased awareness that sometimes lingered beyond the interaction. Engagement was shaped by embodied and curiosity-driven interaction, while prompt effectiveness depended on contextual fit. We conclude with implications for designing technologies that better support informal mindfulness in everyday life. | Koekkoek, Bentvelzen | Short paper |
| 7 | Co-Living Between Courtrooms: Extended Reality as Digital Bridge for Moot Court Training | Moot court training is central to legal education because it helps students develop oral advocacy, confidence, courtroom etiquette, and familiarity with formal legal procedure. Yet physical moot courts are resource intensive and infrequent, and unevenly accessible, particularly for students from under-resourced institutions and remote regions. This paper explores how Extended Reality XR might act as a digital bridge between scarce physical moot court opportunities and more routine, accessible rehearsal. We report an exploratory multi-method study combining expert interviews, semi-structured interviews with students, faculty, lawyers, and judges, passive onsite observations of moot court proceedings, and analysis of moot court documents. The study suggests that the scarcity of moot court opportunities is driven primarily by systemic resource constraints rather than pedagogical preference. Participants also described performance anxiety, limited rehearsal opportunities, and unequal access to training. In response, we designed and prototyped a virtual moot court environment aligned with institutional guidelines used in India. Early qualitative feedback suggests that XR-based moot court environments may support repeated low-stakes practice, broaden access to courtroom rehearsal, and help students become more familiar with courtroom formality and interaction. Positioned within HCI research on immersive learning, embodied interaction, and mixed reality, this work contributes an emerging design space for digitally bridging physical institutions and everyday professional training. | Yadav, Kulkarni, Akhtar | Short paper |
| 8 | “It Might Break but It Has to Work”: Challenging and Comparing Values in Personal Physical Activity Systems | Values play a key role in how we experience technology. Understanding the rationale behind choices between values may help future designers to build systems better aligned with user wants and needs. We present the outcomes of a workshop centred around an adversarial tournament-style ‘values deathmatch’ containing 32 values derived from the literature. These values were discussed in the context of personal exercise systems, with choices based around users’ preferences. Participants debated and chose from randomly paired values, and we analysed their arguments to understand their rationales. We found participants typically prioritised values relating to more effective system outcomes (e.g. Usefulness, Alignment), and gave less weight to values they felt could be as effectively satisfied by themselves, such as Safety or Direction. We highlight how users prioritise values, and we provide rationale-centric framing for discussing values in the context of behaviour change and technologies for supporting physical activity. | Scott, Han, Stawarz | Short paper |
| 9 | Designing and implementing an LLM-Supported Embodied VR Tutor for Smoking Cessation: An Early Prototype | This paper presents an early prototype of an LLM-supported embodied conversational tutor for smoking cessation in immersive virtual reality (VR). The tutor supports spoken interaction with a visible agent and is designed to connect conversational guidance with a planned smoking-trigger context. The dialogue is informed by Motivational Interviewing and Self-Regulated Learning and includes safety constraints to support non-judgemental interaction. The prototype explores how voice-based dialogue and embodiment can be coordinated as a situated human–agent interaction within immersive VR. Early development has highlighted conversational timing and gesture naturalness as areas requiring further refnement. The design provides a basis for studying how embodied conversational support can be delivered within smoking-related situations in immersive VR. The prototype will next be refned through expert and stakeholder feedback before formal user evaluation. | Alshahrani, Cristea | Short paper |
| 10 | Audio-Tactile Paintings for Accessible Interaction in Museums | This paper presents a tactile sonified painting prototype for accessible human-computer interaction in museums. The design responds to a common problem: paintings often depend on sight, while visitors who are blind and visually impaired are mainly offered spoken description or guided support. The prototype turns a selected painting into a touchable and listenable interface. A tactile surface divides the artwork into raised tactile zones, while an ESP32 touch system detects contact and triggers mapped sounds. Spatial audio is used to place these sounds near, far, left, or right, so visitors can build a mental map of colour, shape, texture, mood, and layout. The paper argues that this method is original because it joins tactile modelling, sonification, and spatial listening within museum objects. It also shifts access from passive reception toward active exploration. The paper contributes a practical framework for inclusive museum interaction and future audio-tactile studies. It offers a clear path for museums seeking access tools that are practical and user-led. | Opara, Lewis, Choudhury | Short paper |
| 11 | Making Architectural Decisions Traceable in the Participatory Design of LLM Agents | Participatory design enables HCI researchers to involve stakeholders in shaping interactive systems that reflect their values and contextual needs. In LLM-based conversational agents, however, user experience is shaped not only by visible dialogue or interface features, but also by behind-the-scenes elements such as underlying models and architectures. While participatory design is well-suited to eliciting descriptive user needs, the translation from qualitative design themes to internal agent architectures often remains implicit, distributed across teams, and difficult to trace. This lack of traceability makes it challenging to verify whether the resulting system remains aligned with stakeholder expectations, especially in high-stakes domains like mental health. To address this challenge, this paper explores how to make architectural decisions traceable in the participatory design of LLM-based agents. We present a reflective case situated in the design of an agent for mental health. Through this case, we articulate a translation process shaped by four forms of design judgment: making operations visible, identifying database needs, consolidating overlapping logics, and assigning architectural roles. We show that the same qualitative theme can yield different technical structures, each carrying distinct interactional and ethical consequences. By examining this interpretive translation layer, we illustrate through a reflective case how the movement from participatory insights to agent architecture can be made more traceable. | Cho, Fonseca, Di Simplicio, Calvo | Short paper |
| 12 | How Generative AI Image Generation Aligns with and Shapes Human Mental Images | Although Generative AI (GenAI) image generators are widely used by people to turn an idea into a picture within seconds, the process of generating and iterating over these images based on people’s mental images remains underexplored. We report an exploratory mixed-methods study combining a pre-survey of everyday GenAI image-generation practices (N = 156) with an image-generation task (N = 46) in which participants generated images of an assigned abstract or concrete target word and repeatedly compared the output with their mental image. Participants rather quickly converged on an AI-generated image that matched their mental image, with further iterations refining rather than reshaping it, but this convergence played out differently depending on what was being visualised: abstract concepts left more room for the AI to shape the outcome, while concrete concepts anchored participants to a more stable image that resisted further AI influence. We discuss implications for designing GenAI image-generation interfaces around this distinction to better support the different processes of expressing and updating one’s internal images. | Ghittini, Reicherts | Short paper |
| 13 | RoboCorp: A Structured Corpus and Analysis of News Reports on Robotaxi Incidents | Shared automated vehicles, commonly referred to as robotaxis, are becoming increasingly present in major urban environments. Their deployment introduces new forms of interaction between automated systems, urban infrastructures, and everyday social practices, while a growing number of operational challenges and traffic incidents has raised concerns about how these systems are designed, governed, and made accountable in public space. As an emerging technology, there is currently a lack of systematic collection of these incidents in publicly available datasets. Using a structured, multi-query search and relevance-screening process on the LexisNexis Nexis Uni database, we assembled and coded a corpus of 63 English-language newspaper articles on such incidents, spanning their formative years of deployment from 2016 to 2023, which serves as a starting point for investigating them. We categorise risks, including collisions, operational breakdowns, and situations in which human intervention remains consequential, and examine their influence on perceptions of safety, responsibility, and trust in urban contexts. | Serratore, Schade, Savino, Reicherts, Schöning | Short paper |
| 14 | Hollowing the Collegial: Six Displacements of Collegial Attention in Software Engineering Work | Software teams coordinate through mundane but consequential exchanges: a question asked across a desk, a design rationale worked through aloud, or a judgement made visible in review. Interviews with 33 UK software professionals suggest that LLM use can move some of these exchanges into private worker– model conversations. We examined passages describing changes in consultation, witnessing, explanation, contribution and team learning using qualitative content analysis informed by constant comparison. The analysis yielded six related displacements: witness withdrawal, consultation rerouting, articulation atrophy, reciprocity collapse, attribution obscuration and collective deskilling. Participant accounts show how these shifts appear in practice and connect them to CSCW work on common ground, situated action, boundary objects, practical wisdom and attention. We offer the taxonomy as a sensitising lens rather than a measurement instrument, and close with design directions for keeping model-assisted reasoning inspectable and available to colleagues. | Vishwarupe, Blackwell, Shadbolt, Jirotka, Flechais | Short paper |